A bandgap voltage reference circuit is based on addition of two voltages having equal and opposite temperature coefficients, TC. The first voltage is a base-emitter voltage of a forward biased bipolar transistor. This voltage has a negative TC of about −2.2 mV/C and is usually denoted as a Complementary to Absolute Temperature or CTAT voltage. The second voltage which is Proportional to Absolute Temperature, or a PTAT voltage, is formed by amplifying the voltage difference (ΔVbe) of two forward biased base-emitter junctions of bipolar transistors operating at different current densities. More information on bandgap voltage reference circuits including examples of prior art methodologies can be found in U.S. application Ser. No. 10/375,593 of Stefan Marinca as filed on 27th Feb. 2003, the contents of which are incorporated herein by reference.
It will be understood that for a pair of bipolar transistors operating at collector current densities in a ratio of 1/50 ΔVbe is of the order of about 100 mV at room temperature. As a CTAT voltage is typically of the order of about 700 mV it will be appreciated that the ΔVbe needs to be amplified by an order of about 5, in order to balance the CTAT voltage. However this amplification of the ΔVbe voltage has the effect of including offset voltages into the PTAT voltage and as a result the reference voltage precision may be affected. These errors are greater when the circuitry is implemented using CMOS processes as opposed to similar circuits implemented using bipolar techniques. Such difference in performance can be traced to the fact that in simple CMOS processes only parasitic bipolar transistors are available and amplifiers based on MOS transistors have larger input voltage offsets.
FIG. 1 is an example of a conventional CMOS implemented bandgap voltage reference. Three pMOS transistors M1, M2 and M3 are provided, each having the same aspect ratio of width/length (W/L). A first and second bipolar transistor Q1 and Q2 are provided, with Q2 having a larger emitter area as compared to Q1. As a result Q1 and Q2 operate at a different current density (the emitter currents are the same for both). An amplifier A1, which is coupled to both Q1 and Q2 keeps the two input levels at the same value and as a result a voltage, ΔVbe, is developed across a resistor r1. ΔVbe is of the formΔVbe=(kT/q)(ln(n))  (1)where
k is the Boltzmann constant,
q is the charge on the electron,
T is the operating temperature in Kelvin,
n is the collector current density ratio of the two bipolar transistors.
The voltage reference Vref provided by the circuit can be determined as the base-emitter voltage of Q3 plus the voltage drop over r2.Vref=VbeQ3+(r2/r1)(ΔVbe)  (2)
The scale value of r2/r1 is chosen to be about 5, and as a result the amplifer offset voltage is also amplified by about a factor of 6, as the input offset voltage is gained to the output by a factor of 1+r2/r1. It will therefore be understood that for each 1 mV input voltage offset an error of about 6 mV is reflected into the bandgap reference. One way to reduce this offset sensitivity is to stack the bipolar transistors. The stacking is however limited by the available headroom—most circuits have to operate from an available 2.6V supply voltage and as a result the number of stacks is typically limited to 2 or 3. Therefore although it is know to stack transistors at the input to the amplifier so as to generate a multiple value of ΔVbe, as this is generated across the resistor at the input to the amplifier there is still an offset contribution that is gained by the circuitry.
A further source of error in bandgap voltage reference circuits can be traced to resistor mismatch. As will be evident from an examination of the terms in equation two, any error in the resistor ratio is directly translated into the reference voltage. It would therefore be desirable to minimise this source of error.
Yet another source of error can be traced to what is commonly called “curvature”. This is a second order error component. In a bipolar transistor the base-emitter voltage biased at a PTAT collector current can be given by:
                                          V            be                    ⁡                      (            T            )                          =                                            V                              G                ⁢                                                                  ⁢                0                                      ⁡                          (                              1                -                                  T                                      T                    0                                                              )                                +                                    V                              be                ⁢                                                                  ⁢                0                                      ⁢                          T                              T                0                                              -                                    (                              σ                -                1                            )                        ⁢                          kT              q                        ⁢                          ln              ⁡                              (                                  T                                      T                    0                                                  )                                                                        (        3        )            where:
Vbe(T) is the temperature dependence of the base-emitter voltage for the bipolar transistor at operating temperature,
Vbe0 is the base-emitter voltage for the bipolar transistor at a reference temperature,
VG0 is the bandgap voltage or base-emitter voltage at 0K,
T0 is the reference temperature, and
σ is the saturation current temperature exponent.
The last term of equation three contributes the curvature, and ideally would be minimised.
An example of an implementation of a bandgap circuit in an environment having low operational voltages is provided in U.S. Pat. No. 6,605,987 assigned to Infineon Technologies, AG. This describes the use of lateral transistors for generating a first partial current which has a first temperature dependence. These lateral transistors form an asymmetric input pair of an amplifier and their collector currents drive a second pair of MOS transistors. Due to the coupling of the lateral transistors to the MOS transistors, it is necessary for the MOS transistors to operate at low threshold voltages, which requires implementation using special CMOS processes. Additionally, the circuit requires the use of multiple is resistors, the resistors requiring matching therebetween. Although this circuit is advantageous in environments having only low operational voltages available, it is not suitable for all applications as the requirements for the resistor matching and implementation of the MOS transistors is arduous.
An example of an implementation of a bandgap reference circuit that is specifically designed to reduce the number of resistors utilised is given in U.S. Pat. No. 6,614,209 of Gregoire, Jr. This describes a bandgap reference circuit utilising multiple PTAT sources coupled in series to generate a final PTAT voltage. A current biased base-emitter region of a bipolar transistor is coupled between the final PTAT voltage and an output terminal of the bandgap voltage reference so as to add the base emitter voltage to the final PTAT voltage thereby generating a stable voltage reference. Although this approach enables a reduction in the resistor ratio traditionally used in bandgap circuits it suffers from a number of drawbacks. As the circuit does not provide for curvature correction it is necessary to generate a large ΔVbe (PTAT) in order to balance the base-emiter voltage (CTAT). U.S. Pat. No. 6,614,209 achieves this balancing by using a two-stage architecture as shown in FIG. 5 of the patent. The required PTAT voltage is provided by the combination of an initial PTAT source, referenced as block 510, and a terminal PTAT source, referenced as block 530, both including amplifiers. Due to the configuration of the circuitry, the second amplifier requires a high headroom environment to operate effectively as its two inputs have a common voltage which is about 3 ΔVbe (approximately 330 mV at room temperature) compared to that of the first amplifier. This limits the applications where this bandgap voltage reference can be utilised. Secondly, the requirement for two amplifiers increases the area required and the power supply needed on a die for implementation of such a circuit. Furthermore as there is no curvature correction provided to the reference the precision of the reference voltage provided is limited. Additionally, as two amplifiers are used the contribution to offset and noise is increased.
There is therefore a need to provide a bandgap voltage reference circuit that can be used as a temperature sensor, i.e. is sensitive to temperature fluctuations.